> > > "Linthicum, Tony" <[EMAIL PROTECTED]> writes:
...
> > > * What is the best way to access target level cost information?
> >
> > I'm sure you know that the loop code does this by generating RTL and
> > asking for the cost of it (computation_cost in tree-ssa-loop-ivopts.c).
>
> Which should
> As a first step, to stay on conservative side, it makes sense
> consider the scalar cost of smaller block while calculating scalar cost.
> Note, smaller block may not exist.
I think that this should be considwered quite common.
We should base the weights of the costs of the two blocks on branch
* How do we compare the costs of if-converted vectorized code to
it's scalar counterpart?
o It may be convenient to calculate scalar cost during if-conversion itself.
o It is possible that size of two sibling blocks (Block_A & Block_B)
does not match at the beginning of tree-ssa level if co
Hello,
> "Linthicum, Tony" <[EMAIL PROTECTED]> writes:
>
> > * Would using a tree-level API like estimate_num_insns be superior
> > to using a simple cost of 1 per statment?
>
> For this sort of calculation, I would guess not. I would imagine that
> on most processors the cost of a single
"Linthicum, Tony" <[EMAIL PROTECTED]> writes:
> * Would using a tree-level API like estimate_num_insns be superior
> to using a simple cost of 1 per statment?
For this sort of calculation, I would guess not. I would imagine that
on most processors the cost of a single vector instruction is
Hello all,
Dorit Nuzman and I have been collaborating on a plan for a cost model
for the vectorizer. Included below is an overview of the design for
the initial implementation. We would welcome any input those of you
on the list might have.
Thanks in advance for your help.
Dorit and Tony
--
